Jacob Coxon, a 27-year-old British researcher at Anthropic, resigned around September 8-9, 2026. He warned that the development of self-improving AI could lead to catastrophic outcomes as early as next year.
Colleagues at Anthropic have supported his concerns. Evan Hubinger, Alignment Science Lead, stated there is a greater than 10% chance of AI causing massive human fatalities within the next decade and admitted the company lacks a clear plan for superintelligence alignment.
Researcher Samuel Marks also noted that senior staff at the company are concerned about extinction-level risks. Coxon criticized the industry's focus on competition, citing terms like crunchtime and endgame as factors that prioritize speed over safety.
This departure follows the resignation of Mrinank Sharma, who led the Safeguards Research team, in February 2026. Sharma previously cited external pressures and risks from AI and bioweapons as reasons for his exit.
Coxon highlighted international competition, particularly with Chinese AI labs, as a driver for the current development pace. He argued that the pressure to keep up makes companies reluctant to slow down for safety measures.
